Generative AI Engineer, Regard

• BS in Computer Science or equivalent experience
• 3+ years of professional experience with software development in one or more programming languages (Python preferred)
• 1+ years of professional experience building generative AI products, such as RAGs, agents and chatbots
• Able to participate in on-call operational support for their areas of responsibility
• Able to travel up to 4 weeks a year for company co-working and/or retreat weeks
• Strong verbal and written communication skills

Desirable

• As a Generative AI Engineer at Regard, you’ll work across the full lifecycle of developing and deploying AI-driven features, from ideation and design to prototyping, implementation, evaluation, and iteration
• You’ll collaborate closely with product and clinical teams to build systems that transform medical records into structured insights and clinician-ready documentation
• Your work will center on applying modern LLMs to extract, summarize, normalize, and generate clinical information from diverse electronic health record (EHR) data sources
• This includes developing robust pipelines, running model and prompt-engineering experiments, integrating models into production services, and ensuring outputs remain factual, safe, and clinically aligned
• You’ll directly contribute to high-priority product initiatives, shape new AI capabilities, and advance our LLM platform
• Your work will have a tangible impact on how clinicians understand patient data and how healthcare organizations improve care quality
• Build and refine LLM-powered systems to extract structured medical concepts, diagnoses, medications, labs, and timelines from unstructured records
• Develop generation pipelines that produce clinically accurate drafts of notes (H&P, progress notes, discharge summaries, etc.) from factual inputs
• Design, prototype, and evaluate prompts, agent workflows, and retrieval-augmented generation (RAG) components
• Benchmark LLM systems to evaluate new models and audit accuracy
• Optimize inference cost, latency, and throughput through batching, caching, and model-selection strategies